or

Forgot Password ?

Images steganalysis with unusual scenarios
LIRMM, Montpellier
Marc CHAUMONT
Computer Science and IT Telecommunications Data Analysis 
Full Description:

The topic of this Ph.D. is related to computer science, and more precisely to image processing, pattern recognition, and information technologies.

The steganography / steganalysis can be explained as a game with three participants. The steganographs conventionally called Alice and Bob, want to send a secret message (a binary suite), without being detected. The steganalyst, usually called Eve, observes all the exchanges that take place between Alice and Bob and can stop the communication between Alice and Bob if she finds a secret communication [1].

The Ph.D. student will focus on images steganalysis with unconventional scenarios. He will study scenarios more complex than the usual “clairvoyant” scenario [1], while keeping in mind the very difficult real time traffic analysis scenario.

The scenario “clairvoyant steganalysis” is the most commonly used. This scenario is one of the scenarios introduced in 1983 by Simmons [1] and is similar in spirit to the Kerckhoffs frame-work used in cryptography [2]. It allows evaluating the security of a steganographic algorithm in the worst case.

In this scenario, the steganalyst, Eve, knows the algorithms of insertion and extraction of the message, all the public parameters of these two algorithms (e.g. the number of embedded bits), and she has an images database similar to Alice’s database. The only parameter that Eve does not know is the secret key used by Alice and Bob to insert / extract the message inside images.

Currently, the most efficient steganalysis approaches, for this scenario, use machine learning techniques such as Support Vector Machine [3] or Ensemble Classifiers [4]. This scenario, as well as those that can be derived by eliminating certain assumptions, is based on a wide knowledge of Eve. It is therefore natural to study scenarios more “realistic” where the steganalyst has less knowledge.

As an example, in the “cover-source mismatch” scenario, Eve could not have a perfect knowledge of the type of images used by Alice (their statistical distribution). The Ph.D. student will then have to deal with low-complexity steganalysis in order to face to cover-source mismatch. [5].

Another unusual scenario is when the number of embedded bits is unknown: “message of unknown length” scenario. In this case, using different payloads for the learning (solution proposed by [6]) is severely discussed by [7]. Quantized steganalysis also have to be studied [8].

Finally, the “pooled” scenario occurs when there is “batch steganography” and “pooled steganalysis” [9]. In this scenario, the secret message is transmit not through a single media cover (letter, image, or any digital media) but by spreading it on multiple covers. This scenario is much more difficult for the steganalyst. Recently, Tomas Pevny and Andrew Ker, proposed two papers addressing this problem in practice [10], [11].

The contributions of the PhD student will be multiples. For each studied scenarios, he will realize a complete state of the art, propose a theoretical model, implement it and evaluate it on real datasets. 

Bibliography:
[1] G. J. Simmons, “The prisoners problem and the subliminal channel,” in Advances in Cryptography, CRYPTO, Aug. 1983, pp. 51–67
[2] A. Kerckhoffs, “La Cryptographie Militaire,” Journal des Sciences Militaires, vol. IX, pp. 5-38 Jan. 1883, pp. 161-191, Feb. 1883.
[3] Chih-Chung Chang and Chih-Jen Lin, “LIBSVM: A library for support vector machines,” ACM Transactions on Intelligent Systems and Technology, vol. 2, pp. 27:1–27:27, 2011, Soft-ware available at http://www.csie.ntu.edu.tw/ cjlin/libsvm.
[4] J. Kodovský, J. Fridrich, and V. Holub, “Ensemble classifiers for steganalysis of digital media,” IEEE Transactions on Information Forensics and Security, vol. 7, no. 2, pp. 432–444, 2012.
[5] I. Lubenko and A. D. Ker. Steganalysis with mismatched covers: do simple classifiers help? In Proceedings of the on Multimedia and security, MM&Sec'12, pages 11-18, New York, NY, USA, 2012. ACM.
[6] T. Pevný, « Detecting messages of unknown length », Media Watermarking, Security, and Forensics III, Part of IS&T/SPIE 21th Annual Symposium on Electronic Imaging, SPIE'2011, Volume 7880, San Francisco, California, USA, Feb 2011.
[7] J. Kodovský, J. Fridrich, JPEG-Compatibility Steganalysis Using Block-Histogram of Recompression Artifacts, , 14th Information Hiding Conference, Berkeley, CA, May 15–18, 2012, Springer LNCS vol. 7692, pp. 78-93.[8] J. Kodovský, J. Fridrich, Quantitative Steganalysis Using Rich Models, Proc. SPIE, Electronic Imaging, Media Watermarking, Security, and Forensics XV, vol. 8665, San Francisco, CA, February 3–7, 2013
[9] A. Ker, “Batch steganography and pooled steganalysis,” in Proceedings of the 8th in-ternational conference on Information hiding, Berlin, Heidelberg, 2007, IH’06, pp. 265–281, Springer-Verlag.
[10] A. D. Ker and T. Pevný, “A New Paradigm for Steganalysis via Clustering,” in Media Watermarking, Security, and Forensics XIII, Part of IS&T/SPIE 21th Annual Symposium on Electronic Imaging, SPIE’2011, San Francisco, California, USA, Feb. 2011, vol. 7880, pp. 0U01–0U13.
[11] A. Ker and T. Pevny, “Batch steganography in the real world,” in Proceedings of the on Multimedia and security, New York, NY, USA, 2012, MM&Sec’12, pp. 1–10, ACM.